Cheaper open source alternatives threaten OpenAI

In the ever-changing arenaartificial intelligence (AI), a major challenge is emerging for technology giants like OpenAI, represented by the rise of open source models. This trend, exemplified by innovative companies like Ramp, reveals a strategy where expensive models like OpenAI’s GPT-4 are used as springboards to develop open source alternatives that are more cost-effective and still perform well. While this approach may seem to go against OpenAI’s terms of use (which at this point doesn’t seem to understand the scale of the threat), it is proving to be a strategic pivot for many startups.

Ramp, a fintech startup valued at US$5.8 billion, perfectly illustrates this trend. By leveraging GPT-4’s capabilities to process astronomical amounts of data, Ramp has been able to refine open source models like Mistral AND Llama 2, outperforming GPT-4 in certain tasks at a significantly lower cost. This refinement process, although it may seem like a violation of OpenAI’s terms, is tacitly tolerated, even encouraged, as OpenAI suggests refining its own GPT-3.5 Turbo model with results from GPT-4.

Leave a Comment